Download presentation
Presentation is loading. Please wait.
Published byMiles Daniels Modified over 9 years ago
1
Cambridge Site Report Cambridge Site Report HEP SYSMAN, RAL 10-11 th June 2010 Santanu Das Cavendish Laboratory, Cambridge santanu@hep.phy.cam.ac.uk Santanu Das Cavendish Laboratory, Cambridge santanu@hep.phy.cam.ac.uk
2
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 2 Man Power: For Group John Hill – the main sys-admin Steve Wotton – deputy sys-admin Kavitha Nirmaladevi (half-time) Around 1.2FTE of effort (part of Steve and John) For Grid system Just myself For Group John Hill – the main sys-admin Steve Wotton – deputy sys-admin Kavitha Nirmaladevi (half-time) Around 1.2FTE of effort (part of Steve and John) For Grid system Just myself
3
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 3 Group System: Hardware/OS Around 70 desktops (roughly 2:1 Linux:Windows) Mainly SLC-5.5 and Windows XP (still some SLC4 desktops) Servers are SLC4 at present, serving ~27TB of storage Present Work 18TB of clustered storage to replace ~7TB of the old storage mentioned earlier Migrating from Windows 2000 to Windows 2008 domain Buy ~35TB of storage for LHC n-tuple (and equivalent) Hardware/OS Around 70 desktops (roughly 2:1 Linux:Windows) Mainly SLC-5.5 and Windows XP (still some SLC4 desktops) Servers are SLC4 at present, serving ~27TB of storage Present Work 18TB of clustered storage to replace ~7TB of the old storage mentioned earlier Migrating from Windows 2000 to Windows 2008 domain Buy ~35TB of storage for LHC n-tuple (and equivalent)
4
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 4 Group System: Network Gigabit backbone with a 1Gbps connection onto the University network 10Gbps (??) University connection on to JANET Future Plan Nothing big Buy ~35TB of storage for LHC n-tuple (and equivalent) Traffic levels are rising and we may be forced to consider a upgrade Network Gigabit backbone with a 1Gbps connection onto the University network 10Gbps (??) University connection on to JANET Future Plan Nothing big Buy ~35TB of storage for LHC n-tuple (and equivalent) Traffic levels are rising and we may be forced to consider a upgrade
5
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 5 Grid Status [hardware] : Head nodes Dell PE-1050s (quad-core/8Gb) for CE, SE and UI Worker nodes 33 x Dell PE1950 (2*dual-core 5150 2.66GHz; shared with CamGrid) 4 x Viglan (2*quad-core E5420 @ 2.50Ghz) 4 x SunFires (2*quad-core L5520 @ 2.27GHz) 4 x Dell R410 (2*quad-core E5540 @ 2.53Ghz) Storage 108TB online (~100TB reserved for atlas). ~30TB being used by local project; will be added to the grid soon. Head nodes Dell PE-1050s (quad-core/8Gb) for CE, SE and UI Worker nodes 33 x Dell PE1950 (2*dual-core 5150 2.66GHz; shared with CamGrid) 4 x Viglan (2*quad-core E5420 @ 2.50Ghz) 4 x SunFires (2*quad-core L5520 @ 2.27GHz) 4 x Dell R410 (2*quad-core E5540 @ 2.53Ghz) Storage 108TB online (~100TB reserved for atlas). ~30TB being used by local project; will be added to the grid soon.
6
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 6 Grid Status [middleware] : gLite 3.2 for the WNs. gLite 3.1. for MON and UI. gLite 3.2 for the CE, SE, site-BDII. DPM v1.7.2 on the head node. DPM v1.7.3 on of the DPM disk servers XFS file system for the storage Condor (v7.2.4) is used as the batch system. Supported VOs: Mainly Atlas, LHCb and Camont Additional VO support: Alice, Biomed, Calice, CMS, dteam, euindia, gridpp and obviously ops gLite 3.2 for the WNs. gLite 3.1. for MON and UI. gLite 3.2 for the CE, SE, site-BDII. DPM v1.7.2 on the head node. DPM v1.7.3 on of the DPM disk servers XFS file system for the storage Condor (v7.2.4) is used as the batch system. Supported VOs: Mainly Atlas, LHCb and Camont Additional VO support: Alice, Biomed, Calice, CMS, dteam, euindia, gridpp and obviously ops
7
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 7 Grid System: Network Gigabit backbone with separate 1Gbps connection onto the University network All the WNs are on the gigabit-network 10Gbps (??) University connection on to JANET Future Plan Nothing big Buy ~35TB of storage for LHC n-tuple (and equivalent) But traffic levels are rising and we may be forced to consider a upgrade Network Gigabit backbone with separate 1Gbps connection onto the University network All the WNs are on the gigabit-network 10Gbps (??) University connection on to JANET Future Plan Nothing big Buy ~35TB of storage for LHC n-tuple (and equivalent) But traffic levels are rising and we may be forced to consider a upgrade
8
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 8 Grid System [issues]: Network Middleware is too buggy for Condor No proper/practical support, yet All the previously written scripts are almost no longer maintained Most of the “info-provider” scripts rewritten /modified locally Every new release breaks the condor-glite integration Cannot use yaim on CE Spending too much time on fixing glite scripts rather trying new things. GridPP4 money, of course. Network Middleware is too buggy for Condor No proper/practical support, yet All the previously written scripts are almost no longer maintained Most of the “info-provider” scripts rewritten /modified locally Every new release breaks the condor-glite integration Cannot use yaim on CE Spending too much time on fixing glite scripts rather trying new things. GridPP4 money, of course.
9
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 9 Grid System [plans]: Network Upgrade Condor More job-slots and disk space Condor on Cream-CE Install Scas,glexec Network Upgrade Condor More job-slots and disk space Condor on Cream-CE Install Scas,glexec
10
11/06/2010santanu@hep.phy.cam.ac.uk HEP, Cavendish Laboratory 10 Questions??
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.